Multi-Task Deep Learning Network With Power-State Layer Activation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Multi-task deep learning networks require complex circuits and suffer from power waste and failure in low power supply conditions, leading to inefficient operation of electronic apparatuses.

Innovation Solution

A multi-task deep learning network design with shared and task-specific layers that activate based on power supply state, optimizing power usage by enabling only necessary layers for task execution.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multiple independent networks are used to perform different tasks, then task execution capability is improved, but power consumption increases and device complexity increases

Engineering Contradiction:

Improvetask execution capabilityVSAvoidpower consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent merges multiple independent task-specific networks into a unified multi-task deep learning network architecture where task-specific layers are integrated under a shared framework. This allows the system to execute multiple tasks through a single network structure rather than running separate independent networks, thereby reducing overall power consumption while maintaining task execution capability.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent implements dynamic activation of task-specific layers based on power supply state. The system can dynamically adjust which task-specific layers are activated - enabling full functionality when power is abundant and selectively disabling less critical tasks when power is limited. This dynamic adaptation resolves the contradiction by making task execution capability flexible rather than fixed.

Inventive Principle:

Principle #15Dynamics

2Adaptability or versatility

If multiple independent networks are used to perform different tasks, then task execution capability is improved, but device complexity increases

Engineering Contradiction:

Improvetask execution capabilityVSAvoidcircuit complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent combines multiple task-specific processing paths into a single integrated network architecture. Instead of implementing separate independent networks for each task, the system uses one unified network with shared components and task-specific layers that can be selectively activated. This merging approach maintains the ability to perform multiple tasks while significantly reducing circuit complexity compared to having fully independent networks for each task.

Inventive Principle:

Principle #5Merging (Combining)

3Adaptability or versatility

If all task-specific layers are activated, then full functionality is achieved, but power consumption increases

Engineering Contradiction:

ImprovefunctionalityVSAvoidpower waste

Core Design Contradiction:

Adaptability or versatilityVSLoss of energy

Solution Approach 1:

The patent implements a dynamic power management mechanism that adjusts the activation state of task-specific layers based on real-time power supply conditions. When power supply is sufficient, all task-specific layers are activated to provide full functionality. When power supply becomes limited, the system dynamically deactivates less critical task-specific layers while maintaining activation of essential tasks, thereby preventing power waste while preserving core functionality.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the operational parameters of the network by adjusting which task-specific layers are active based on power supply state. This parameter change allows the system to transition between different operational modes - from full functionality with all layers active to power-saving mode with selective layer activation - thereby optimizing the balance between functionality and power consumption.

Inventive Principle:

Principle #35Parameter changes

4Loss of energy

If task-specific layers are activated in stages according to power supply state, then power efficiency is improved, but system complexity increases

Engineering Contradiction:

Improvepower efficiencyVSAvoidnetwork architecture complexity

Core Design Contradiction:

Loss of energyVSDevice complexity

Solution Approach 1:

The patent segments the network into a shared base layer and multiple task-specific layers. This segmentation allows the system to activate only the necessary task-specific layers while keeping the shared layer continuously active. The segmentation approach enables power-efficient staged activation without requiring complete independence between task components, thereby managing system complexity in a structured way.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12602587B2Multi-task deep learning network and generation method thereof

Publication Date: 2026.04.14 ALI CORP

AI summary

A multi-task deep learning network and a generation method thereof are provided. The generation method of a multi-task deep learning network includes: building at least one shared layer, where the shared layer is configured to receive a plurality of pieces of input information and generate a plurality of pieces of processed feature information; building a plurality of groups of task-specific layers, wherein the groups of task-specific layers compute and generate a plurality of groups of output information corresponding to a plurality of different tasks according to the pieces of processed feature information; and activating at least one of the groups of task-specific layers in stages according to a power supply state of an electronic apparatus.